The Compelling Business Imperative of Open Source Software

Fred Goldstein,  May 2003


...Open Source is beneficial because its participants are socially optimizing the value of their work. By building on a base of work that is open to the public, Open Source developers reduce the amount of repetitive work that each has to do...


...The program often does what its author needs it to do, not what other potential users might expect or want it to do. That key marketing function, product management, simply doesn't exist.....

One of the principles of network design and optimization is the distinction between "selfish" and "social" optimization. The protocol that a computer uses to communicate on a shared network can be optimized to benefit a specific computer, to the expense of others who share the network. For example, a single node on a shared Ethernet might use shorter-than-average backoff times, so that it often goes ahead of others, or it might cheat on TCP's "slow start" congestion control policy. In such situations, the overall throughput of the network declines; if too many computers practice selfish optimization, the network is even subject to congestion collapse, a kind of massive traffic pile-up when essentially nothing gets through. Social optimization benefits all computers equally; it prevents congestion collapse by having all users take only their own fair share of the resources.

A similar principle seems to be applicable to the world of software as well. While software is not a closed network with a finite amount of bandwidth to be divided up, it is nonetheless subject to both selfish and social optimizations. This may explain the natural growth in the related "Open Source" and "Free Software" movements, which are producing "free" software that competes, often successfully, with proprietary products. Open Source is beneficial because its participants are socially optimizing the value of their work. By building on a base of work that is open to the public, Open Source developers reduce the amount of repetitive work that each has to do, allowing each developer to focus on new capabilities and product differentiators.

The software developer community is fundamentally divided into two camps. A minority of developers is engaged in the production of software as a product which is licensed for a fee. The largest example of this, of course, is Microsoft. Other firms such as Oracle, Symantec and Corel specialize in this; it is also a major revenue stream for computer companies such as IBM and Hewlett-Packard. For these companies, open source software represents a threat. Linux now represents the largest threat to Microsoft's dominance of the desktop; it and the BSDs already command a large share of the server market.

The larger number of developers, though, are not engaged in selling software; they are instead writing software to solve specific problems on behalf of their employers. Such in-house developers are the major beneficiaries of free software. Not only does it reduce the price they have to pay, compared to proprietary systems, but it also provides them with a code base to build on to. They have to reinvent fewer wheels, and are thus free to concentrate more closely on their specific tasks. In-house programs and even in-house modifications to GPL code are not subject to that license's strict sharing requirements. Thus a company can use Linux, BSD or even open source Windows tools to produce process control, financial, scientific, and other internal applications.

Desktops have proven less amenable

Today, the vast majority of desktop computers are using a proprietary operating system. Microsoft still has over 90% of the market with its various flavors of Windows. Apple is valiantly struggling along, its MacOS X based on a mix of an open-source BSD-related kernel (Darwin) supporting its proprietary Macintosh-specific features. (MacOS X is thus able to support both MacOS software and many generic Unix applications.)  Sun's Solaris and other Unix variants have niche markets among high-end users. But Linux is passing MacOS in desktop share and gaining converts daily.

Linux did not become the leading Unix desktop on technical merit alone; certainly if technical quality were so important, then OS/2 would have overpowered Windows 3.1. Windows benefitted from the network effect: More users attracted more developers, which meant more software, which meant more users. This positive-feedback loop resembles other natural monopolies, in which the largest market share equates to the lowest unit cost. Competing with a natural monopoly is tricky. It generally begins at the margins, where the monopolist's economy of scale can be overcome by economy of specialization. That was the case with OS/2, for instance, which attracted users drawn to its superior stability.  Just not very many. Linux is just now getting that network effect: It has coalesced into "the" alternative desktop OS for Intel-class hardware. It has the attention of the press. It has the attention of Microsoft, which has tried several strategies to disparage it. So more developers are focusing on Linux, creating a software base that attracts a growing number of users, etc.

Of course the Linux ecosystem does not resemble Microsoft's. Most of the important Linux-based software projects are themselves open source. Linux users are, in large part, a parsimonious bunch, who are usually happiest with software that costs no more than the downloadable version of their favorite distribution. Indeed, the mere act of installing a distribution like SuSe or Mandrake Linux has a surreal quality to it: With so many free software packages included with the distribution, it's rather like going to a software store with a blank check. This tends to limit the attractiveness of Linux as a platform for commercial software developers. While some vendors, like Oracle, have had success selling onto Linux, others have found it hard to compete with free. This makes it somewhat less than likely that your favorite Windows package will be found on Linux.

Not that the free application software on Linux is always the equal of its proprietary counterparts. The Linux market is still largely composed of geeks, so the Linux world to some extent is reminiscent of the DEC world of two decades ago -- systems built by geeks for geeks. Neither was totally successful in their attempts to move into other markets. Linux application software is often deficient in the user interface, not to mention documentation -- all too often, the authors assume that the source code is documentation enough! And with no money on the line, there's little incentive to do that last 10% of the work that typically takes most of the effort. The program often does what its author needs it to do, not what other potential users might expect or want it to do. That key marketing function, product management, simply doesn't exist. Again, the fact that the source code is open provides an excuse, albeit a ridiculously weak one: If you don't like it, code it yourself. Such an attitude is less common now than it was a few years ago, but it is still all too present in the Linux community.

Fortunately, a few companies are making serious efforts to provide "horizontal" applications to the general user base. Sun Microsystems, for instance, turned over the bulk of its Star Office package, a reasonable, if not quite polished, alternative to Microsoft Office, to an open source group, Thus there is a free "OOo" office suite and a somewhat larger, still proprietary StarOffice suite built on top of it. Ximian produces Evolution, a clone of Microsoft Outlook. Well, a 90% clone -- while it intentionally omits certain well-known Evolution features, such as its virus susceptibility and propagation. But a few desirable features, like "delete from server after x days", are also missing, having fallen below the line.

A small but growing number of organizations, especially in the public sector, where money can be scarce, have begun to migrate desktops to open source, especially Linux. And with the rapidly improving quality of the Linux desktop, it is graining at least some credibility as an alternative to Windows for power users.

Proprietary software still has its place.  Some applications are very support-intensive; for example, an antivirus program is almost useless if it is not updated for a month.  And some vertical market applications, like telephone billing, are extremely complex and have narrow target marketplaces. Commercial software provides a viable "buy, don't make" alternative.  Of course, such commercial software may itself benefit from open-source components or subsystems.

The biggest action will be behind the scenes, with servers and company-specific vertical applications. By providing a free and growing baseline upon which to build application-specific and user-specific code, open source software has created its own compelling business case. Companies should not be afraid to adopt it.

 "Free" in this case has two different meanings, both of which fall within the same context of social optimization.  "Free as in speech" is the mantra of the Free Software Foundation, whose General Public License (GPL) requires that programs which incorporate any GPL code are themselves subject to the GPL.  Most of the Linux system is under the GPL. "Free as in beer" describes the Berkeley-style licenses used by various BSD systems.  These can not only be used and distributed for no charge; they can also be incorprated freely into proprietary programs. Much of the Internet's core server software is under BSD licenses.

OS/2's eventual failure in mass markets, however, was less one of marketing than one of the monopolist's pressure.  Microsoft reportedly told IBM that the price they would have to pay for o.e.m. Windows 95 licenses would be far higher than what its competitors paid, unless they agreed to limit OS/2 sales.  The difference in Windows license fees alone was larger than total OS/2 revenues.

I suggest that Apple has pursued a suicidal strategy here by tying sales of MacOS to its own hardware.  While this has helped their Macintosh hardware line, it has crippled sales of the higher-margin software, especially given the low incremental cost of software sales and the network effects of higher market share. It is of course purely a matter of speculation to guess how much profit Apple could have made selling MacOS for PC-class commodity systems.